@techreport{oai:ipsj.ixsq.nii.ac.jp:00217848,
 author = {李, 越東 and Li, Yuedong},
 issue = {35},
 month = {May},
 note = {リモートセンシングは元々，従来の深層学習モデルでよく用いられるようなオブジェクト検出や分類のためではないだろう．著者は従来の YOLOv4 ネットワーク構造に基づいて，小さな目標に対する 104＊104 の特徴検出層を追加し，csSE モジュールを追加し，活性化関数は LeakyRelu を MISH 関数に設定されることで，衛星画像中の小さな目標の検出能力を向上させることがわかった．結果は，DIOR データセットで mAP@0.5 は約 9.12％ の改善が見られた．, It is a difficult problem how to make traditional neural network algorithm show good adaptability to the typical target detection of remote sensing image in the field of remote sensing. In probing the latest YOLOv4 core idea, network structure and algorithm, the network structure is firstly improved by adding 104×104 feature layer scale and embedding csSE module. Then, use the Mish activation function to replace the original activation function Leaky ReLU to obtain better generalization, and the typical target detection algorithm performance of YOLOv4 is improved for remote sensing image. Finally, it is verified by designing contrast experiment. The results showed that the mAP@0.5 both increased by 9.12% on the DIOR test set.},
 title = {レイヤーとアテンションを追加したYOLO-v4による小さな目標に頑健な物体検出},
 year = {2022}
}